In this complex task, a multi-hop prompt instructs the agent to go through a 8 newly imported PDFs (stored locally on my desktop) to find important quotes, claims, and points made by the authors that could be used in a paper im working on.
After finding these points throughout the 8 files, the agent should create a new document that cites findings in-text, with verifiable click through.
With Gemini 3.1 Flash, Ubik Agents analyzed 8 PDFs, found 49 usable quotes, and generated a 1000 word annotated guide that explains its findings with in-text citations that verifiably link to cited points of information across documents inside my workspace in 6 minutes (video 2x speed).
Ubik is free until 2026! While we figure out how to make Ubik the best research AI tool possible, we value our users' feedback instead of churning quickly in these early days. This design philosophy has really helped us develop with the user in mind. That being said we plan to have a Free, Pro ($12-$15 + pay as you go with model usage), and Enterprise. If you have any feedback, please let me know! Happy you enjoyed Ubik. You can reach me at [email protected] any time :)
I've built a similar platform with searching access to arXiv and Semantic Scholar; the only difference is that our agents can highlight text in the paper down to the line level. In our testing, Gemini struggles compared to Sonnet four or Opus 4. We found that without agentic highlighting, there wasn't much difference in output quality or utility (meaning, references saved, generating with citation, or even quote gathering is still hard without actual PDF interactivity). I'd love your feedback on https://www.ubik.studio (use academic search)
Our academic search uses ArXiv and Semantic Scholar, so when using the academic agent, you can convert open-access papers into AI-ready docs directly from the databases. <33
The video is sped up at 2.5x, and the actual time spent was around 7 minutes.
We are now setting up an Eval suite and hope to run evals on our agents within the coming month. We are almost finished with the entire local app.
Ubik is an AI studio with two agents: Fiig (built for research workflows (searching, analyzing, and writing)and Precog (Just like Fiig, but Web search-based instead of using academic databases).
If this looks like a helpful tool for you, I'd love to hear your feedback!
After finding these points throughout the 8 files, the agent should create a new document that cites findings in-text, with verifiable click through.
With Gemini 3.1 Flash, Ubik Agents analyzed 8 PDFs, found 49 usable quotes, and generated a 1000 word annotated guide that explains its findings with in-text citations that verifiably link to cited points of information across documents inside my workspace in 6 minutes (video 2x speed).
Try Ubik: www.ubik.studio/download